CN110991452B - Parking space frame detection method, device, equipment and readable storage medium - Google Patents
Parking space frame detection method, device, equipment and readable storage medium Download PDFInfo
- Publication number
- CN110991452B CN110991452B CN201911223321.7A CN201911223321A CN110991452B CN 110991452 B CN110991452 B CN 110991452B CN 201911223321 A CN201911223321 A CN 201911223321A CN 110991452 B CN110991452 B CN 110991452B
- Authority
- CN
- China
- Prior art keywords
- parking space
- image
- space frame
- parking
- initial positioning
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/24—Aligning, centring, orientation detection or correction of the image
- G06V10/245—Aligning, centring, orientation detection or correction of the image by locating a pattern; Special marks for positioning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/155—Segmentation; Edge detection involving morphological operators
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20036—Morphological image processing
Abstract
The application discloses a parking space frame detection method, which is used for realizing parking space frame detection based on deep learning, avoiding equipment installation and maintenance cost caused by invasive detection, simultaneously removing irrelevant background through semantic segmentation network prediction, extracting the approximate area of a parking space frame to obtain a binarized parking space area, simplifying regression tasks, effectively avoiding the influence of insufficient illumination, shielding pedestrians and sundries, and the like, further accurately positioning an initial positioning image through a multi-point regression network, increasing fitting capacity on the parking space shape through the multi-point regression network, realizing high-precision parking space frame detection, and improving detection accuracy. The application also discloses a parking space frame detection device, a device and a readable storage medium, which have the beneficial effects.
Description
Technical Field
The application relates to the field of internet of things, in particular to a parking space frame detection method, device and equipment and a readable storage medium.
Background
In recent years, the average holding capacity of automobiles in China rises year by year, and parking space resources are limited, so that the problem of difficult parking is caused, and the problems of difficult parking space, uneven utilization of the parking space resources and the like are mainly represented. Therefore, the method has important significance in real-time monitoring of the parking spaces, improving management of parking spaces, roadside parking monitoring, parking guidance, improving utilization efficiency of the parking spaces and the like.
In the process of automatically managing a parking space, parking is an important step in detection. The parking space detection methods commonly used at present are mainly classified into invasive detection and non-invasive detection. The invasive detection mainly comprises the steps of detecting parking space states through induction coils and microwaves, and the non-invasive detection mainly comprises the steps of video image processing, radar image processing, passive infrared and the like. The sensor and buried line based invasive parking space detection method is relatively complex in construction, is easily interfered by various factors, is unstable in accuracy and lacks of safety; the non-invasive parking space detection method based on the image is often subjected to the problems of insufficient light, difference of ground background of the parking space, shielding of the parking space, interference of ground stains and the like, and the conventional detection method is difficult to solve.
Disclosure of Invention
The application aims to provide a parking space frame detection method which can realize high-precision parking space frame detection and improve the detection accuracy; another object of the present application is to provide a parking space frame detecting apparatus, a device and a readable storage medium.
In order to solve the technical problems, the application provides a parking space frame detection method, which comprises the following steps:
acquiring a parking area image;
the parking space frame area initial detection is carried out on the parking area image through a semantic segmentation network, so that an initial positioning image is obtained;
and accurately positioning the initial positioning image through a multipoint regression network to obtain an accurate positioning frame wire, and taking the accurate positioning frame wire as a parking space frame wire to be output.
Optionally, before the accurate positioning of the initial positioning image by the multipoint regression network, the method further comprises:
performing morphological post-processing on the initial positioning image to obtain an optimized image;
wherein the morphological post-processing comprises: removing spots in small areas in the initial positioning image, and/or connecting the separated parking space areas in the initial positioning image, and/or removing false detection in the initial positioning image;
the initial positioning image is correspondingly accurately positioned through a multipoint regression network, which comprises the following steps: and accurately positioning the optimized image through a multipoint regression network.
Optionally, before the accurate positioning of the initial positioning image by the multipoint regression network, the method further comprises:
performing region outward expansion on the initial positioning image according to a preset image outward expansion proportion to obtain an outward expansion image;
the initial positioning image is correspondingly accurately positioned through a multipoint regression network, which comprises the following steps: and accurately positioning the outward expansion image through a multipoint regression network.
Optionally, the performing area expansion on the initial positioning image according to a preset image expansion ratio includes:
determining the minimum bounding box of the split parking space frame area in the initial positioning image;
expanding the upper and lower boundaries of the bounding box by 0.05 times of the height and expanding the left and right boundaries by 0.1 times of the width to obtain a new bounding box;
and carrying out image interception according to the new bounding box, and taking the intercepted image as the expansion image.
Optionally, the multipoint regression network is: based on 4 vertexes, the long side of the parking space frame is 22 linear points, and the short side is 10 linear points which are used as regression points.
Optionally, the acquiring the parking area image includes:
acquiring a parking space picture shot by a parking lot monitoring camera;
and carrying out format preprocessing on the parking space picture, and taking the preprocessed image as the parking area image.
Optionally, the accurately positioning the initial positioning image through the multipoint regression network includes:
extracting image features through a multipoint regression network to carry out regression, so as to obtain perspective transformation coefficients;
performing perspective transformation processing on the standard parking space frame according to the perspective transformation coefficient to obtain multi-point position coordinates of the parking space frame;
and generating a parking space frame line according to the parking space frame multi-point position coordinates.
The application discloses a parking space frame detection device, which comprises:
an image acquisition unit configured to acquire a parking area image;
the semantic segmentation unit is used for carrying out initial detection on the parking space frame area of the parking area image through a semantic segmentation network to obtain an initial positioning image;
the multipoint regression unit is used for accurately positioning the initial positioning image through the multipoint regression network to obtain an accurate positioning frame wire, and taking the accurate positioning frame wire as a parking space frame wire to be output.
The application discloses a parking space frame detection device, which comprises:
a memory for storing a program;
and the processor is used for realizing the step of the parking space frame detection method when executing the program.
The application discloses a readable storage medium, wherein a program is stored on the readable storage medium, and the program realizes the steps of the parking space frame detection method when being executed by a processor.
The parking space frame detection method provided by the application comprises the following steps: acquiring a parking area image; the parking space frame area initial detection is carried out on the parking area image through a semantic segmentation network, so that an initial positioning image is obtained; and accurately positioning the initial positioning image through a multipoint regression network to obtain an accurate positioning frame wire, and taking the accurate positioning frame wire as a parking space frame wire to be output. According to the method, parking space frame detection is achieved based on deep learning, equipment installation and maintenance cost caused by invasive detection is avoided, irrelevant background rejection is carried out through semantic segmentation network prediction, the general area of a parking space frame is extracted, a binarized parking space area is obtained, regression tasks are simplified, influences such as insufficient illumination, shielding of pedestrians and sundries are effectively avoided, an initial positioning image is accurately positioned through a multipoint regression network, fitting capacity to a parking space shape is improved through the multipoint regression network, high-accuracy parking space frame detection can be achieved, and detection accuracy is improved.
The application also discloses a parking space frame detection device, a device and a readable storage medium, which have the beneficial effects and are not repeated here.
Drawings
In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings that are required to be used in the embodiments or the description of the prior art will be briefly described below, and it is obvious that the drawings in the following description are only embodiments of the present application, and that other drawings can be obtained according to the provided drawings without inventive effort for a person skilled in the art.
Fig. 1 is a flowchart of a parking space frame detection method provided by an embodiment of the present application;
FIG. 2 is a schematic view of a parking area according to an embodiment of the present application;
FIG. 3 is a schematic view of another parking area according to an embodiment of the present application;
FIG. 4 is a schematic diagram of a semantic segmentation network segmentation result according to an embodiment of the present application;
fig. 5 is a schematic view of an extracted parking space frame area according to an embodiment of the present application;
FIG. 6 is a schematic diagram of another semantic segmentation network segmentation result according to an embodiment of the present application;
fig. 7 is a schematic view of another extracted parking space frame area according to an embodiment of the present application;
fig. 8 is a schematic diagram of a symmetrical network structure according to an embodiment of the present application;
FIG. 9 is a schematic diagram of a labeling sample according to an embodiment of the present application;
FIG. 10 is a schematic diagram of another labeling sample according to an embodiment of the present application;
FIG. 11 is a schematic illustration of a multi-point annotation provided by an embodiment of the present application;
FIG. 12 is a schematic illustration of another multi-point annotation provided by an embodiment of the present application;
fig. 13 is a schematic diagram of a parking space frame prediction process based on a multipoint regression network according to an embodiment of the present application;
fig. 14 is a block diagram of a parking space frame detection device according to an embodiment of the present application;
fig. 15 is a block diagram of a parking space frame detection device according to an embodiment of the present application;
fig. 16 is a schematic structural diagram of a parking space frame detection device according to an embodiment of the present application.
Detailed Description
The core of the application is to provide a parking space frame detection method, which has simple device, can effectively avoid the influence of insufficient illumination, shielding of pedestrians and sundries, and has higher accuracy; the application further provides a parking space frame detection device, a device and a readable storage medium.
For the purpose of making the objects, technical solutions and advantages of the embodiments of the present application more apparent, the technical solutions of the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application, and it is apparent that the described embodiments are some embodiments of the present application, but not all embodiments of the present application. All other embodiments, which can be made by those skilled in the art based on the embodiments of the application without making any inventive effort, are intended to be within the scope of the application.
The application provides a parking space frame detection method which is applicable to, but not limited to, indoor parking lots and roadside parking application scenes. Referring to fig. 1, fig. 1 is a flowchart of a parking space frame detection method provided in the present embodiment; the method mainly comprises the following steps:
step s110, obtaining a parking area image;
the parking area image refers to a parking area image containing a parking space frame, in the embodiment, the image source is not limited, the image acquisition can be carried out by associating with a camera acquired in real time, and the parking space frame detection can also be carried out by receiving a historical image. Optionally, a process of acquiring an image of a parking area specifically includes the steps of:
acquiring a parking space picture shot by a parking lot monitoring camera;
and carrying out format preprocessing on the parking space picture, and taking the preprocessed image as a parking area image.
According to the method for acquiring the image of the traffic area, provided by the embodiment, the parking space frame can be directly monitored in real time by directly correlating with the parking lot monitoring camera, and the real-time parking space frame detection efficiency can be improved. The preprocessing process may generally include adjustment of image size and pixel values, and uniform image size, such as size to 320×160, to facilitate uniform and accurate analysis of images of different sizes; the normalization processing of the pixel values is convenient for unified and accurate analysis of images with different pixel values, for example, the normalization of the image pixel divided by 255 to 0-1 is performed by taking the preprocessing means as an example in the embodiment, and of course, if the obtained original image is in a unified format, a unified size, a unified pixel value, a unified shooting angle and the like, the obtained parking space image can also be directly detected for a parking space frame, which is not limited herein.
Step 120, carrying out parking space frame area initial detection on the parking area image through a semantic segmentation network to obtain an initial positioning image;
the initial positioning image is a parking space frame area segmented by a semantic segmentation network. In the embodiment, a binarized parking space area is obtained through semantic segmentation network prediction, the background is removed, the approximate area of a parking space frame is extracted, and a regression task is simplified.
After the initial detection of the parking space frame through the semantic segmentation network model, as shown in fig. 2 and 3, the segmentation result of the semantic segmentation network corresponding to fig. 2 is shown in fig. 4, the segmentation result of the semantic segmentation network corresponding to fig. 3 is shown in fig. 5, the segmentation result of the semantic segmentation network corresponding to fig. 6 is shown in fig. 6, the segmentation result of the semantic segmentation network corresponding to the extracted parking space frame (i.e. the obtained initial positioning image) is shown in fig. 7, and it can be found that the removal of the irrelevant background area and the rough positioning of the parking space frame can be realized through the semantic segmentation network.
It should be noted that, the semantic segmentation network structure invoked in the embodiment is not limited, and may refer to the structure setting in the conventional image segmentation method, fig. 8 is a schematic diagram of a symmetrical network structure, in the embodiment, only the above network structure is taken as an example, other network structures are not described herein, and the corresponding adjustment of the network structure may be performed according to the actual image segmentation requirement. In addition, the semantic segmentation network is obtained based on the parking lot parking space image training of the marked parking space frame, the training process of the network is not limited in the embodiment, and the training process of the traditional network model can be referred to.
To further understand, a training process of a network model is described in this embodiment, which is specifically as follows: and gathering 2789 parking space pictures (the number of the pictures is not limited) of the parking space in different scenes according to the target image shot by the parking space monitoring camera, and marking the positions of the parking space frames. When making labels, the mask size can completely comprise the parking space frame (as shown by the labels on the right side of two vehicles in fig. 9 and the labels between two vehicles in fig. 10), namely, the function of the semantic segmentation network is to realize rough positioning of the parking space frame area, set a training set and a testing set of the required parking space frame, and label the target area and the label. And optimizing the weight parameters of the semantic segmentation network by using the training samples and the labels to obtain an optimal segmentation model. And then the optimal segmentation model can be called to carry out initial detection of the parking space frame.
The semantic segmentation network is used for detecting the parking space frame area, a parking space frame initial positioning image excluding a large number of background area images can be obtained, the obtained initial positioning image can be directly input into the multipoint regression network to accurately position the parking space frame, and the binary parking space area is obtained through the prediction of the semantic segmentation network due to the interference of light rays, noise, ground stains and the like in a scene, so that the problems that the segmentation is discontinuous, the cavities exist and the like are easy to occur, the fuzzy positioning during the multipoint regression is avoided, the positioning accuracy is influenced, and preferably, the morphological post-processing can be further carried out on the initial positioning image before the accurate positioning is carried out on the initial positioning image through the multipoint regression network, so that an optimized image is obtained.
Removing spots in small areas by morphological dilation and etching operations in order to eliminate areas of wrong segmentation; connecting the separated parking space areas; considering that the parking space frame is generally large in size in the image, partial virtual detection can be further removed through the mode of limiting the size and the length-width ratio of the minimum bounding box of the area and the like. The image optimization process may include: spots in small areas in the initial positioning image are removed, the separated parking space areas in the initial positioning image are connected, the virtual inspection in the initial positioning image is removed, one or more of the virtual inspection in the initial positioning image can be selected, other morphological post-processing means can be further added, the method is not limited, and the method is only described by taking the morphological post-processing means as an example in the embodiment.
Because some parking space frame areas can be cut off when parking space frame area initial detection is carried out on a parking area image through a semantic segmentation network, the cut-off parking space frame areas cannot be embodied in subsequent parking space detection, and the precision of parking space frame detection is affected, the parking space frame areas which are not detected can be properly expanded, and the precision of parking space frame detection is improved. In this embodiment, the specific expansion ratio is not limited, and specifically, performing area expansion on the initial positioning image according to the preset image expansion ratio may specifically include: determining the minimum bounding box of the split parking space frame area in the initial positioning image; expanding the upper and lower boundaries of the bounding box by 0.05 times of the height and expanding the left and right boundaries by 0.1 times of the width to obtain a new bounding box; and carrying out image interception according to the new bounding box, and sending the intercepted parking space frame region RGB image into a multipoint regression network.
It should be noted that, the two image processing means (morphological post-processing and image expansion) described in this step may not be executed, one of them may be selected for execution, or both may be executed, and when both are executed, the execution sequence is not limited. It should be noted that, the back-end image processing manner may be set correspondingly in the front-end network training process, for example, when the initial detection is performed, the morphological post-processing is performed, and then the accurate detection is performed, and the corresponding execution sequence should be kept as much as possible in the multipoint regression network training process, so as to ensure the optimal image detection effect.
And step 130, accurately positioning the initial positioning image through a multipoint regression network to obtain an accurate positioning frame wire, and taking the accurate positioning frame wire as a parking space frame wire to be output.
The multipoint regression network can accurately detect the parking space frame line, and in order to obtain the parking space shapes under various installation angles, the fitting capacity of the multipoint regression network on the parking space shapes is increased.
The network structure of the invoked multipoint regression network model in the embodiment is not limited herein, and can be set according to the network structure in the traditional method, and the general parking space frame is considered to be quadrilateral, and the parking space frame line presented in the image of the shooting angle problem is trapezoid, and only the regression of four vertexes is considered to easily generate larger errors, so that the patent adopts the multipoint mode to regress the parking space line. Preferably, the multipoint regression network is specifically: based on 4 vertexes, the long side of the parking space frame is 22 linear points, and the short side is 10 linear points which are used as regression points. Based on four vertexes, 22 points are linearly generated on the long side of the parking space frame, 10 points are generated on the short side, 68 points are taken as regression points, the obtained multi-point labeling schematic diagrams are shown in fig. 11 and 12, and the multi-point labeling can avoid excessive calculation while ensuring the accuracy.
In order to avoid the error caused by the detection of the situation, the process of accurately positioning the initial positioning image through the multi-point regression network specifically comprises the following steps:
step S131, extracting image features through a multipoint regression network to carry out regression, and obtaining perspective transformation coefficients;
step s132, performing perspective transformation processing on the standard parking space frame according to the perspective transformation coefficient to obtain multi-point position coordinates of the parking space frame;
and step S133, generating a parking space frame line according to the multi-point position coordinates of the parking space frame.
Based on the analysis, the method adopts a mode of regression perspective transformation matrix and scaling coefficient to carry out multi-point regression, and the mode enhances the relevance among the multiple points. The parameters of the CNN network structure for performing multipoint regression positioning according to the above manner are shown in the following table, and it should be noted that in this embodiment, only the above network structure parameters are described as examples, and other structure parameter settings based on the present application can refer to the following table, which is not described herein.
Filters, kernel, stride in the table represents the number and size of convolution kernels of the convolution layer and the sliding step length of the convolution kernels, and output represents the size of the output characteristic diagram of the convolution layer.
Fig. 13 shows a schematic diagram of a parking space frame prediction process based on the above multi-point regression network, firstly, the parking space area image resize is reduced to 64 x 64, the translation, perspective transformation coefficient and scaling coefficient are obtained by regression through CNN extraction image features, and then the final multi-point position coordinates of the parking space frame can be obtained by perspective transformation of the standard parking space frame (with the size of 64 x 128) through translation, rotation, scaling and the like. Correspondingly, before the multipoint regression prediction is performed by adopting the process, when the multipoint regression network is trained, a training sample needs to be preprocessed, wherein the training sample comprises images resize to 64 x 64, the size is uniform, the image pixels are divided by 255 and normalized to be between 0 and 1, and in addition, four vertexes of a parking space line need to be marked on the basis of a parking space frame area image roughly positioned by a semantic segmentation network.
The multi-point regression model is used for predicting the to-be-processed parking space frame area, so that the accurate positioning of the parking space frame line can be obtained, and the accurate detection of the parking space frame is realized.
Based on the above description, the parking space frame detection method provided by the embodiment realizes the detection of the parking space frame based on the deep learning, avoids the equipment installation and maintenance cost caused by the invasive detection, simultaneously performs irrelevant background rejection through semantic segmentation network prediction, extracts the approximate area of the parking space frame, obtains the binarized parking space area, simplifies the regression task, effectively avoids the influence of insufficient illumination, shielding of pedestrians and sundries, and the like, further performs accurate positioning on the initial positioning image through the multipoint regression network, increases the fitting capability on the parking space shape through the multipoint regression network, can realize the detection of the parking space frame with high accuracy, and improves the detection accuracy.
In order to enhance the understanding of the specific implementation flow of the parking space frame detection method described in the above embodiment, the description of the parking space frame detection flow based on the above embodiment in this embodiment is described in this embodiment, and other parking space frame detection based on the above embodiment may refer to the description of this embodiment, which is not repeated herein. The method specifically comprises the following steps:
1. and acquiring field images, and manufacturing a training set and a testing set.
The labeling manner is as shown in the above embodiment, and will not be described herein.
The semantic segmentation network data set mainly comprises the steps that a labeling mask completely covers a parking space area and is used as a label.
The multipoint regression network data set is marked, four vertexes of a parking space are marked first, and 68 points are uniformly generated on four sides. The fact that the shapes of the parking space frames in the images are different is considered due to the fact that the installation positions of the cameras are different in an actual scene. In order to obtain the parking space shape under various installation angles, the fitting capacity of a multipoint regression network on the parking space shape is increased, perspective transformation processing is carried out on each piece of data of the training set, and each image generates 6 rotating images with different perspective angles to serve as an enhanced training set.
2. Semantic segmentation network training is performed based on the sample data.
And sending the enhanced training set into a semantic segmentation network, predicting to obtain a parking space area, calculating softmax loss with label, and adjusting network convolution layer parameters through back propagation.
3. And performing multipoint regression network training based on the sample data.
The dataset for the multi-point regression is derived from the predicted outcome of the semantic segmentation network. The enhanced training set obtains a predicted parking space area through a semantic segmentation network, the height of the upper and lower boundaries of the area is expanded by 0.05 times, the width of the left and right boundaries is expanded by 0.1 times, the obtained result is used as a rough parking space frame area image, and then the predicted parking space area is sent into a multipoint regression network for training. After the perspective transformation parameters are predicted by multi-point regression, the labeling parking space frames are subjected to perspective transformation to obtain parking space frame line point coordinates, euclidean Loss is calculated with the labeling point coordinates, and the network convolution layer parameters are adjusted through back propagation.
4. And (5) predicting a parking space frame.
The method comprises the steps of predicting the approximate parking space area of an image to be detected through a semantic segmentation network, removing spots of small areas through morphological operation, and connecting the isolated parking space areas. And then expanding the parking space area, predicting perspective transformation parameters through a multipoint regression network, and finally, carrying out perspective transformation on the standard parking space frame to obtain the final position of the parking space frame line point.
In the parking space frame detection process based on deep learning, coarse positioning of a parking space frame area is achieved through a semantic segmentation network, accurate parking space line positioning is achieved through a multipoint regression network, and parking space frame shape regression is achieved through prediction perspective transformation parameters in the multipoint regression network. The device that this realization process adopted is simple, and can effectively avoid illumination inadequacy, pedestrian, debris to shelter from etc. influence, has higher rate of accuracy.
Referring to fig. 14, fig. 14 is a block diagram of a parking space frame detecting device according to the present embodiment; the apparatus may include: an image acquisition unit 210, a semantic segmentation unit 220, and a multipoint regression unit 230. The parking space frame detection device provided by the embodiment can be mutually compared with the parking space frame detection method introduced in the embodiment.
Wherein, the image acquisition unit 210 is mainly used for acquiring a parking area image;
the semantic segmentation unit 220 is mainly used for carrying out initial detection on the parking space frame area of the parking area image through a semantic segmentation network to obtain an initial positioning image;
the multipoint regression unit 230 is mainly used for accurately positioning the initial positioning image through the multipoint regression network to obtain an accurate positioning frame wire, and taking the accurate positioning frame wire as a parking space frame wire to be output.
The parking space frame detection device provided by the embodiment is simple in structure, can effectively avoid the influence of insufficient illumination, shielding of pedestrians and sundries and the like, and has higher accuracy rate
Referring to fig. 15, fig. 15 is a block diagram of a parking space frame detecting apparatus according to the present embodiment; the apparatus may include: memory 300 and processor 310. The parking space frame detection equipment can refer to the introduction of the parking space frame detection method.
The memory 300 is mainly used for storing programs;
the processor 310 is mainly used for implementing the steps of the above parking space frame detection method when executing a program.
Referring to fig. 16, a schematic structural diagram of a parking space frame detection device according to the present embodiment may generate relatively large differences due to different configurations or performances, and may include one or more processors (central processing units, CPU) 322 (e.g., one or more processors) and a memory 332, one or more storage media 330 (e.g., one or more mass storage devices) storing application programs 342 or data 344. Wherein the memory 332 and the storage medium 330 may be transitory or persistent. The program stored on the storage medium 330 may include one or more modules (not shown), each of which may include a series of instruction operations in the data processing apparatus. Still further, the central processor 322 may be configured to communicate with the storage medium 330 and execute a series of instruction operations in the storage medium 330 on the parking space frame detecting device 301.
The parking space frame detection device 301 may also include one or more power supplies 326, one or more wired or wireless network interfaces 350, one or more input/output interfaces 358, and/or one or more operating systems 341, such as Windows ServerTM, mac OS XTM, unixTM, linuxTM, freeBSDTM, and the like.
The steps in the parking space frame detection method described in fig. 1 above may be implemented by the structure of the parking space frame detection apparatus described in this embodiment.
The embodiment discloses a readable storage medium, on which a program is stored, and the steps of a parking space frame detection method are implemented when the program is executed by a processor, where the parking space frame detection method can refer to the corresponding embodiment of fig. 1, and will not be described herein.
In the description, each embodiment is described in a progressive manner, and each embodiment is mainly described by the differences from other embodiments, so that the same similar parts among the embodiments are mutually referred. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant points refer to the description of the method section.
Those of skill would further appreciate that the various illustrative elements and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware, computer software, or combinations of both, and that the various illustrative elements and steps are described above generally in terms of functionality in order to clearly illustrate the interchangeability of hardware and software. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the solution. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present application.
The steps of a method or algorithm described in connection with the embodiments disclosed herein may be embodied directly in hardware, in a software module executed by a processor, or in a combination of the two. The software modules may be disposed in Random Access Memory (RAM), memory, read Only Memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.
The method, the device, the equipment and the readable storage medium for detecting the parking space frame provided by the application are described in detail. The principles and embodiments of the present application have been described herein with reference to specific examples, the description of which is intended only to facilitate an understanding of the method of the present application and its core ideas. It should be noted that it will be apparent to those skilled in the art that various modifications and adaptations of the application can be made without departing from the principles of the application and these modifications and adaptations are intended to be within the scope of the application as defined in the following claims.
Claims (9)
1. The parking space frame detection method is characterized by comprising the following steps of:
acquiring a parking area image;
the parking space frame area initial detection is carried out on the parking area image through a semantic segmentation network, so that an initial positioning image is obtained;
accurately positioning the initial positioning image through a multipoint regression network to obtain an accurate positioning frame wire, and taking the accurate positioning frame wire as a parking space frame wire to be output;
the accurately positioning the initial positioning image through the multipoint regression network comprises the following steps:
extracting image features of the initial positioning image through the multipoint regression network to carry out regression so as to obtain a perspective transformation coefficient;
performing perspective transformation processing on the standard parking space frame according to the perspective transformation coefficient to obtain multi-point position coordinates of the parking space frame;
and generating a parking space frame line according to the parking space frame multi-point position coordinates.
2. The parking space frame detection method according to claim 1, further comprising, before the initial positioning image is accurately positioned by a multipoint regression network:
performing morphological post-processing on the initial positioning image to obtain an optimized image;
wherein the morphological post-processing comprises: removing spots in small areas in the initial positioning image, and/or connecting the separated parking space areas in the initial positioning image, and/or removing false detection in the initial positioning image;
the initial positioning image is correspondingly accurately positioned through a multipoint regression network, which comprises the following steps: and accurately positioning the optimized image through a multipoint regression network.
3. The parking space frame detection method according to claim 1, further comprising, before the initial positioning image is accurately positioned by a multipoint regression network:
performing region outward expansion on the initial positioning image according to a preset image outward expansion proportion to obtain an outward expansion image;
the initial positioning image is correspondingly accurately positioned through a multipoint regression network, which comprises the following steps: and accurately positioning the outward expansion image through a multipoint regression network.
4. The parking space frame detection method of claim 3, wherein the performing area expansion on the initial positioning image according to a preset image expansion ratio comprises:
determining the minimum bounding box of the split parking space frame area in the initial positioning image;
expanding the upper and lower boundaries of the bounding box by 0.05 times of the height and expanding the left and right boundaries by 0.1 times of the width to obtain a new bounding box;
and carrying out image interception according to the new bounding box, and taking the intercepted image as the expansion image.
5. The parking space frame detection method according to claim 1, wherein the multipoint regression network is: based on 4 vertexes, the long side of the parking space frame is 22 linear points, and the short side is 10 linear points which are used as regression points.
6. The parking space frame detection method according to claim 1, wherein the acquiring the parking area image includes:
acquiring a parking space picture shot by a parking lot monitoring camera;
and carrying out format preprocessing on the parking space picture, and taking the preprocessed image as the parking area image.
7. Parking stall frame detection device, its characterized in that includes:
an image acquisition unit configured to acquire a parking area image;
the semantic segmentation unit is used for carrying out initial detection on the parking space frame area of the parking area image through a semantic segmentation network to obtain an initial positioning image;
the multi-point regression unit is used for accurately positioning the initial positioning image through a multi-point regression network to obtain an accurate positioning frame wire, and taking the accurate positioning frame wire as a parking space frame wire to be output;
the accurately positioning the initial positioning image through the multipoint regression network comprises the following steps:
extracting image features of the initial positioning image through the multipoint regression network to carry out regression so as to obtain a perspective transformation coefficient;
performing perspective transformation processing on the standard parking space frame according to the perspective transformation coefficient to obtain multi-point position coordinates of the parking space frame;
and generating a parking space frame line according to the parking space frame multi-point position coordinates.
8. Parking stall frame check out test set, its characterized in that includes:
a memory for storing a program;
the processor is configured to implement the steps of the parking space frame detection method according to any one of claims 1 to 6 when the program is executed.
9. A readable storage medium, wherein a program is stored on the readable storage medium, which when executed by a processor, implements the steps of the parking space frame detection method according to any one of claims 1 to 6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911223321.7A CN110991452B (en) | 2019-12-03 | 2019-12-03 | Parking space frame detection method, device, equipment and readable storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911223321.7A CN110991452B (en) | 2019-12-03 | 2019-12-03 | Parking space frame detection method, device, equipment and readable storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110991452A CN110991452A (en) | 2020-04-10 |
| CN110991452B true CN110991452B (en) | 2023-09-19 |
Family
ID=70089719
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911223321.7A Active CN110991452B (en) | 2019-12-03 | 2019-12-03 | Parking space frame detection method, device, equipment and readable storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110991452B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113449648A (en) * | 2021-06-30 | 2021-09-28 | 北京纵目安驰智能科技有限公司 | Method, system, equipment and computer readable storage medium for detecting indicator line |
| CN113963571B (en) * | 2021-10-28 | 2023-10-17 | 深圳市捷顺科技实业股份有限公司 | Method and device for processing identification event of vehicle entering and exiting parking lot |
| CN114255584B (en) * | 2021-12-20 | 2023-04-07 | 济南博观智能科技有限公司 | Positioning method and system for parking vehicle, storage medium and electronic equipment |
| WO2023216251A1 (en) * | 2022-05-13 | 2023-11-16 | 华为技术有限公司 | Map generation method, model training method, readable medium, and electronic device |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101145240A (en) * | 2007-11-09 | 2008-03-19 | 王海燕 | Camera image road multiple-point high precision calibration method |
| CN105809106A (en) * | 2016-02-23 | 2016-07-27 | 北京理工大学 | Vehicle formation following detection method based on machine vision |
| CN109410635A (en) * | 2018-10-26 | 2019-03-01 | 北京智芯原动科技有限公司 | A kind of trackside parking space state recognition methods and device |
| CN110210350A (en) * | 2019-05-22 | 2019-09-06 | 北京理工大学 | A kind of quick parking space detection method based on deep learning |
| CN110443225A (en) * | 2019-08-15 | 2019-11-12 | 安徽半问科技有限公司 | A kind of actual situation Lane detection method and device thereof based on statistics of pixel eigenvalue |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10078790B2 (en) * | 2017-02-16 | 2018-09-18 | Honda Motor Co., Ltd. | Systems for generating parking maps and methods thereof |
-
2019
- 2019-12-03 CN CN201911223321.7A patent/CN110991452B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101145240A (en) * | 2007-11-09 | 2008-03-19 | 王海燕 | Camera image road multiple-point high precision calibration method |
| CN105809106A (en) * | 2016-02-23 | 2016-07-27 | 北京理工大学 | Vehicle formation following detection method based on machine vision |
| CN109410635A (en) * | 2018-10-26 | 2019-03-01 | 北京智芯原动科技有限公司 | A kind of trackside parking space state recognition methods and device |
| CN110210350A (en) * | 2019-05-22 | 2019-09-06 | 北京理工大学 | A kind of quick parking space detection method based on deep learning |
| CN110443225A (en) * | 2019-08-15 | 2019-11-12 | 安徽半问科技有限公司 | A kind of actual situation Lane detection method and device thereof based on statistics of pixel eigenvalue |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110991452A (en) | 2020-04-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110991452B (en) | Parking space frame detection method, device, equipment and readable storage medium | |
| CN110390267B (en) | Mountain landscape building extraction method and device based on high-resolution remote sensing image | |
| CN107341453B (en) | Lane line extraction method and device | |
| JP6336117B2 (en) | Building height calculation method, apparatus and storage medium | |
| CN111080693A (en) | Robot autonomous classification grabbing method based on YOLOv3 | |
| US7620266B2 (en) | Robust and efficient foreground analysis for real-time video surveillance | |
| JP6488083B2 (en) | Hybrid method and system of video and vision based access control for parking space occupancy determination | |
| US9576199B2 (en) | Method and device for calculating number and moving direction of pedestrians | |
| US20120321134A1 (en) | Face tracking method and device | |
| KR102074073B1 (en) | Method for detecting vehicles and apparatus using the same | |
| CN102706274B (en) | System for accurately positioning mechanical part by machine vision in industrially-structured scene | |
| CN109685045A (en) | A kind of Moving Targets Based on Video Streams tracking and system | |
| CN110910445B (en) | Object size detection method, device, detection equipment and storage medium | |
| CN106127205A (en) | A kind of recognition methods of the digital instrument image being applicable to indoor track machine people | |
| AU2020272936B2 (en) | Methods and systems for crack detection using a fully convolutional network | |
| WO2023273011A1 (en) | Method, apparatus and device for detecting object thrown from height, and computer storage medium | |
| CN109784145B (en) | Target detection method based on depth map and storage medium | |
| CN111695373A (en) | Zebra crossing positioning method, system, medium and device | |
| CN111881748B (en) | Lane line visual identification method and system based on VBAI platform modeling | |
| CN113112479A (en) | Progressive target detection method and device based on key block extraction | |
| WO2022045877A1 (en) | A system and method for identifying occupancy of parking lots | |
| CN111325198B (en) | Video object feature extraction method and device, and video object matching method and device | |
| CN107067411B (en) | Mean-shift tracking method combined with dense features | |
| CN116310889A (en) | Unmanned aerial vehicle environment perception data processing method, control terminal and storage medium | |
| CN116188826A (en) | Template matching method and device under complex illumination condition |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |